This is a nice catch, Kant is indeed inconsistent in his use of "pure". Below I am quoting from the Guyer's 1998 translation of the Critique.
In Section I Kant first distinguishes between empirical and a priori, then among the latter, between relative and absolute, and, finally, among the absolute, between pure and impure propositions/judgments. The "pure absolute" means that not even the concepts within are derived from experience. Alas, this scope of "pure" turns out to be empty outside of mathematics, and the Critique is mostly about application of a priori cognition to empirical matters. What follows is a common phenomenon in language use: when a term becomes idle in some context (here, use of understanding outside of mathematics) its meaning is shifted to make it useful again.
So in the second section Kant redefines the "pure", without announcing it, by adding "strictly universal" to what he previously called "absolute". At the start of B5 he states as much:
"Now it is easy to show that in human cognition there actually are such necessary and in the strictest sense universal, thus pure a priori
However, "strictly universal" just means "in such a way that no exception at all is allowed to be possible" (B4), which certainly does not rule out the use of concepts derived from experience. He then first gives an example from mathematics, which qualifies as originally "pure", but only concerns forms of sensibility, and then predicates his second example on "if one would have one from the commonest use of the understanding". Well, the "commonest use of the understanding" is unifying sensible experience, so it can not possibly provide anything "pure" in the original sense.
"If one wants an example from the sciences, one need only
look at all the propositions of mathematics; if one would have one
from the commonest use of the understanding, the proposition that
every alteration must have a cause will do; indeed in the latter the very
concept of a cause so obviously contains the concept of a necessity of
connection with an effect and a strict universality of rule that it would
be entirely lost if one sought, as Hume did, to derive it from a frequent
association of that which happens with that which precedes and a habit..."
What's more, it turns out that "strict universality" does not really add anything to "necessity", and hence "absoluteness", except the ease of verification. In other words, the original pure/impure distinction within the absolute a priori is essentially erased.
"Necessity and strict universality are therefore
secure indications of an a priori cognition, and also belong together inseparably. But since in their use it is sometimes easier to show the empirical limitation in judgments than the contingency in them, or is often more plausible to show the unrestricted universality that we ascribe to a judgment than its necessity, it is advisable to employ separately these two criteria, each of which is in itself infallible."
This is confirmed by the opening sentence of Section II, which plainly ignores all the finer distinctions and simply opposes "pure" (absolute a priori) to "empirical":"At issue here is a mark by means of which we can securely distinguish a pure cognition from an empirical one".